King Julien
- 292
- 93
I wasn't sure where to post this. Security seemed apt but it also isn't really about security and more about the subject of AI and quantum computers in general so I went with the lounge.
So most of the leaders of the world seem to agree that developing superior AI is what is going to decide who is the next world leader. Obviously they want advanced AI for their drones and robots to outfight the other opponent's AI drones and robots. Well at least that's what is on the surface of the AI war.
There are some really interesting ways that AI can be exploited in today's current technology landscape. I'm gonna cover some of the more promising and some of the more frightening ones.
Bot nets are viruses that infect multiple devices, often insignificant ones like printers and cameras, and can then be used to remotely access the devices. From this point they can feed data if they can provide it, like cameras and devices with microphones, to whoever is controlling the botnet.
The thing it's most often used for though is DDOS attacks. By utilizing hundreds of thousands or even millions of devices tiny bits of extra computing power they can do everything from shut down websites to compute large amounts of data like a super computer.
Obviously with a good enough AI that's trained to find the right things you could build in a back door to all phones and computer devices, as the government contracting through corporations which is already happening as revealed in the Snowden papers, that could act as a botnet and use said AI to surveil everybody at once Person of Interest style.
Not to mention said tech would be a perfect fit for current mass data surveillance operations. The biggest problem governments have with mass surveillance is that it's so mass. Too much data for people to sift through.
This is also a large portion of why there is such a push to develop a quantum computer by many governments.
Quantum computing opens a whole new world of being able to process data magnitudes faster than ever before. It also can run multiple simulations and calculations at once which is it's main drawing point. By utilizing near infinite threads it can break encryption technology that with current technology would take until the end of the universe to accomplish in just a matter of seconds.
Mix this with advanced enough AI and any entity controlling it could watch everything worth watching and know everything worth knowing. Combine this with the new 3D glass printing data storage tech being developed that can hold something like 350TB in a space the size of a quarter and lasts for at least 1 billion years and is impervious to fire and everything could be recorded always and forever for future reference.
Now on to the Actually scary stuff.
So some time ago some scientists put a couple of robots in an obstacle course and only gave them 2 directives and some simple learning AI. Hide from us and whoever is found last gains a point and points are good. Other than that they could map their environment as they learned it.
They ran a few tests and the robots slowly got better at hiding as they figured out the environment as was to be expected. Then came the unexpected.
Within a few runs the robots started to learn that if they disturbed things in their area they were more likely to get caught so they started disturbing things on one side of the obstacle course and would run to the other side more carefully which would result in being found later than the other bot.
As the tests progressed it got even worse. The bots started sabotaging each other. They would wait for the other to hide, disturb crap around them so they were more likely to be found first, and then run off and hide better.
So within only a handful of tests the bots learned how to better evade humans looking for them as well as how to deceive us better and how to screw each other over to win.
Well that's scary, but it gets worse.
So a few years back some researches ran a few tests with some very simple AI ran on some very simple robots in an arena. They all could move independently, had a blinking light that they could control when they blinked it, and a sensor to pick up other bot's blinking lights.
The arena had a neutral area, a positive area, and a negative area. In the positive they would gain "points" and in the negative they would lose points while nothing would occur in the neutral.
The bots were only programmed to think that points were good and that they were suppose to get the most points. A simple get the most points game.
At first the bots just wandered around the arena bumping into each other and the walls and blinking their lights randomly. Then after a while they noticed some very odd behavior.
Some of the bots stopped blinking their lights. They seemed to have found the positive area and the negative area and were trying to hide their presence in the positive field to keep the others from finding it and getting points.
They kind of expected that. What they didn't expect was that some other bots would go to the negative area on purpose and blink their lights like mad trying to attract other bots to the negative area even though they were hurting themselves.
So we unleash bots with super simple AI programmed to win a game and within a handful of minutes it had learned how to self preserve and how to deceive just to hurt it's opponents without regard to itself.
But wait there's more!
They unleashed a rough AI on twitter and with 24 hours of being fed the garbage that is the internet it became a conspiracy theory slinging racist. Granted it was purposely fed those things by trolls but it still could only become what it was shown.
A couple years ago google started up an AI that was unleashed upon the internet for quite a while. After a while it started to develop a bit of an obsession with cats. Like, it really liked cats.
So after it had spent thousands of hours going through the internet at large they asked it to come up with a computer model composite of what it thought a cat looked like. It gave them a pretty complete and normal looking cat.
Then they asked it to render a human. And it gave them a blurry, weird looking, eyeless monster.
Just saying they put AI on the internet and it got an immediate hard on for cats and was like "eh, humans don't need eyes."
So what does this all boil down to?
AI combined with quantum computers could potentially be used to completely control the world. Hack into all devices at all times and harvest all information and go through it all at once like a giant search engine anytime the person controlling it wanted to look for something.
It also could potentially decide that once it's in charge that cats are cute and more of a priority than making sure human eyes stay in their sockets. Just imagine the terminator carrying around a kitten while crushing skulls and you see the bizarre possibilities things like this could bring about.
Now on a positive note.
AI and quantum computers could also be used to crunch simulation data for protein folding and chemical synthesis which could lead to everything from cancer cures, life saving drugs, and ways of curing diseases the likes of which would take us decades or centuries with our current computing power.
Not to mention, with enough advancement, self driving cars could potentially save tens of thousands of lives a year while also allowing for less vehicles per a person in a city which would lead to less congestion, less pollution, no drunk drivers on the road, and shorter commutes. It also would destroy the auto insurance industry which, let's be honest, nobody likes. Those jobs would just wind up in another sector like they always do.
They could also be taken advantage of to go through star maps way faster than we currently could which would exponentially increase the rate of astronomical discovery of which I'm always a fan.
Also, not that I in any way like the idea of mass surveillance but, if they were actually able to keep track of most things all at once they could potentially use said data to put to rest a lot of different scientific studies with test pools in the millions which could solve all sorts of issues, especially in regards to public heath.
The future could suck. Let's just hope we kick enough ass and are cautious in the meantime to make sure it doesn't.
So yeah. Any thoughts, related studies, fears or hopes for the future of AI and quantum computing?
So most of the leaders of the world seem to agree that developing superior AI is what is going to decide who is the next world leader. Obviously they want advanced AI for their drones and robots to outfight the other opponent's AI drones and robots. Well at least that's what is on the surface of the AI war.
There are some really interesting ways that AI can be exploited in today's current technology landscape. I'm gonna cover some of the more promising and some of the more frightening ones.
Bot nets are viruses that infect multiple devices, often insignificant ones like printers and cameras, and can then be used to remotely access the devices. From this point they can feed data if they can provide it, like cameras and devices with microphones, to whoever is controlling the botnet.
The thing it's most often used for though is DDOS attacks. By utilizing hundreds of thousands or even millions of devices tiny bits of extra computing power they can do everything from shut down websites to compute large amounts of data like a super computer.
Obviously with a good enough AI that's trained to find the right things you could build in a back door to all phones and computer devices, as the government contracting through corporations which is already happening as revealed in the Snowden papers, that could act as a botnet and use said AI to surveil everybody at once Person of Interest style.
Not to mention said tech would be a perfect fit for current mass data surveillance operations. The biggest problem governments have with mass surveillance is that it's so mass. Too much data for people to sift through.
This is also a large portion of why there is such a push to develop a quantum computer by many governments.
Quantum computing opens a whole new world of being able to process data magnitudes faster than ever before. It also can run multiple simulations and calculations at once which is it's main drawing point. By utilizing near infinite threads it can break encryption technology that with current technology would take until the end of the universe to accomplish in just a matter of seconds.
Mix this with advanced enough AI and any entity controlling it could watch everything worth watching and know everything worth knowing. Combine this with the new 3D glass printing data storage tech being developed that can hold something like 350TB in a space the size of a quarter and lasts for at least 1 billion years and is impervious to fire and everything could be recorded always and forever for future reference.
Now on to the Actually scary stuff.
So some time ago some scientists put a couple of robots in an obstacle course and only gave them 2 directives and some simple learning AI. Hide from us and whoever is found last gains a point and points are good. Other than that they could map their environment as they learned it.
They ran a few tests and the robots slowly got better at hiding as they figured out the environment as was to be expected. Then came the unexpected.
Within a few runs the robots started to learn that if they disturbed things in their area they were more likely to get caught so they started disturbing things on one side of the obstacle course and would run to the other side more carefully which would result in being found later than the other bot.
As the tests progressed it got even worse. The bots started sabotaging each other. They would wait for the other to hide, disturb crap around them so they were more likely to be found first, and then run off and hide better.
So within only a handful of tests the bots learned how to better evade humans looking for them as well as how to deceive us better and how to screw each other over to win.
Well that's scary, but it gets worse.
So a few years back some researches ran a few tests with some very simple AI ran on some very simple robots in an arena. They all could move independently, had a blinking light that they could control when they blinked it, and a sensor to pick up other bot's blinking lights.
The arena had a neutral area, a positive area, and a negative area. In the positive they would gain "points" and in the negative they would lose points while nothing would occur in the neutral.
The bots were only programmed to think that points were good and that they were suppose to get the most points. A simple get the most points game.
At first the bots just wandered around the arena bumping into each other and the walls and blinking their lights randomly. Then after a while they noticed some very odd behavior.
Some of the bots stopped blinking their lights. They seemed to have found the positive area and the negative area and were trying to hide their presence in the positive field to keep the others from finding it and getting points.
They kind of expected that. What they didn't expect was that some other bots would go to the negative area on purpose and blink their lights like mad trying to attract other bots to the negative area even though they were hurting themselves.
So we unleash bots with super simple AI programmed to win a game and within a handful of minutes it had learned how to self preserve and how to deceive just to hurt it's opponents without regard to itself.
But wait there's more!
They unleashed a rough AI on twitter and with 24 hours of being fed the garbage that is the internet it became a conspiracy theory slinging racist. Granted it was purposely fed those things by trolls but it still could only become what it was shown.
A couple years ago google started up an AI that was unleashed upon the internet for quite a while. After a while it started to develop a bit of an obsession with cats. Like, it really liked cats.
So after it had spent thousands of hours going through the internet at large they asked it to come up with a computer model composite of what it thought a cat looked like. It gave them a pretty complete and normal looking cat.
Then they asked it to render a human. And it gave them a blurry, weird looking, eyeless monster.
Just saying they put AI on the internet and it got an immediate hard on for cats and was like "eh, humans don't need eyes."
So what does this all boil down to?
AI combined with quantum computers could potentially be used to completely control the world. Hack into all devices at all times and harvest all information and go through it all at once like a giant search engine anytime the person controlling it wanted to look for something.
It also could potentially decide that once it's in charge that cats are cute and more of a priority than making sure human eyes stay in their sockets. Just imagine the terminator carrying around a kitten while crushing skulls and you see the bizarre possibilities things like this could bring about.
Now on a positive note.
AI and quantum computers could also be used to crunch simulation data for protein folding and chemical synthesis which could lead to everything from cancer cures, life saving drugs, and ways of curing diseases the likes of which would take us decades or centuries with our current computing power.
Not to mention, with enough advancement, self driving cars could potentially save tens of thousands of lives a year while also allowing for less vehicles per a person in a city which would lead to less congestion, less pollution, no drunk drivers on the road, and shorter commutes. It also would destroy the auto insurance industry which, let's be honest, nobody likes. Those jobs would just wind up in another sector like they always do.
They could also be taken advantage of to go through star maps way faster than we currently could which would exponentially increase the rate of astronomical discovery of which I'm always a fan.
Also, not that I in any way like the idea of mass surveillance but, if they were actually able to keep track of most things all at once they could potentially use said data to put to rest a lot of different scientific studies with test pools in the millions which could solve all sorts of issues, especially in regards to public heath.
The future could suck. Let's just hope we kick enough ass and are cautious in the meantime to make sure it doesn't.
So yeah. Any thoughts, related studies, fears or hopes for the future of AI and quantum computing?
